Wall Street Prompt · Marketing Cloud

THE PORTFOLIO MONITORING AGENT Build a watcher, not an oracle. It should never tell you what to buy. It should tell you what changed against your thesis, with the receipt. AI watches. You judge. (swipe)

Everyone wants the same thing: "An agent that tells me what to buy." Wrong job. That agent makes you obey. A good agent makes you read. The real job: notice when reality moves against the thesis you wrote down.

START WITH THE THESIS, NOT THE TOOL Write down what you are betting on. -> 3 to 5 assumptions, in plain words -> Example shape: "margins hold," "this segment grows," "debt is manageable" -> Each one is testable No written thesis = nothing to monitor.

DEFINE WHAT BREAKS EACH ASSUMPTION The agent needs triggers, not vibes. -> Margins -> a new 10-Q or earnings print -> Growth -> segment detail in the filing -> Debt -> maturity schedule, new issuance -> Each assumption maps to a source it lives in

WATCH THE SOURCE, NOT THE PRICE Price is noise. Disclosure is evidence. -> Monitor filings, calls, and official prints -> Not the ticker, not the chatter -> A price drop is a question -> A changed disclosure is an answer

EVERY ALERT CARRIES A SOURCE TRAIL This is the whole discipline. -> Document + page + date on every alert -> A link back to the primary source -> No source trail = the alert does not send -> You can verify it in under a minute, or it is useless

WHAT THE AGENT IS NOT DOING -> Not deciding if the change is good -> Not telling you what it is worth -> Not telling you to act The agent compressed the watching. You still own the assumptions. You still own the thesis. AI watches. You judge. Save this. Build it on your holdings. Educational only. Not advice.

A portfolio agent that tells you what to buy is a liability, not a tool. It trains you to obey instead of read.

Build the opposite: a watcher whose only job is to notice when reality moves against the thesis you already wrote down, and to hand you the source so you can judge it.

The build, in order:

1. Write the thesis. 3 to 5 testable assumptions you are actually betting on. 2. Map the triggers. Which event in which document could break each assumption. 3. Watch the source, not the price. Filings, calls, and official prints are evidence. The ticker is noise. 4. Demand a source trail. Document, page, and date on every alert, or it does not send. 5. Classify, do not conclude. Confirms thesis, threatens thesis, or unclear. "Unclear" goes on the follow-up list.

Notice what the agent never does: decide if the change is good, what it is worth, or whether you act. That stays with you.

The common mistake is building an oracle. The failure mode is you stop verifying. The discipline is a watcher that always shows its work.

Human judges. AI builds the watcher.

Save this one and build it against your own holdings.

Educational content only. Not investment advice, and not a recommendation to buy, sell, or hold any security. Wall Street Prompt. Always verify against the primary source.

Build a Portfolio Monitoring Agent (That Watches Your Thesis, Not the Price)

Slides

Caption

Design notes

CTA